Parameter drift in EAs: the setting that changed without anyone noticing
An EA's behavior depends on its parameters, and parameters can change unexpectedly. A platform update may reset inputs. A template change may overwrite settings. Copying a chart can preserve old values. Occasionally, a mistaken edit goes unnoticed for days.
Symptoms include a sudden change in trade frequency, size, or stop distance. Sometimes performance suffers, and the cause is only found after trawling through logs.
Prevent it by storing parameter sets as files and checking them at startup. Code can log all inputs on initialization, creating a record. Comparing the logged values against the intended settings catches discrepancies.
Version control is useful too. Keeping dated copies of the EA and its settings makes it possible to trace when a change occurred.
For multi-account or multi-chart setups, a central configuration reduces the chance of inconsistency.
A brief weekly audit of running EAs and their parameters takes minutes and avoids unpleasant surprises.
What process ensures that EA settings stay as intended across restarts and updates?
Parameter drift in EAs: the setting that changed without anyone noticing
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: Parameter drift in EAs: the setting that changed without anyone noticing
After a Sunday terminal update my take-profit input silently reset from nine pips to four, which was the default baked into the new compiled file. Trade frequency looked similar so I did not notice until Wednesday, and I gave back twenty-seven pips that should have been banked at the old target. A copied chart template had kept an older magic number as well, which made the journal harder to read until I lined the inputs up against a note I had written the previous month.
The process I use now is dull and it works. Every accepted set file lives in a dated folder, the expert prints every input to the Experts log on startup, and I keep a one-page sheet of intended values. Each Sunday I spend four minutes comparing the log line to the sheet before I walk away.
Two updates later that checklist caught a lot size of zero point twelve against the intended zero point ten after I had cloned a chart. It takes longer to describe than to do, and it is the only reason I trust a restart.
The process I use now is dull and it works. Every accepted set file lives in a dated folder, the expert prints every input to the Experts log on startup, and I keep a one-page sheet of intended values. Each Sunday I spend four minutes comparing the log line to the sheet before I walk away.
Two updates later that checklist caught a lot size of zero point twelve against the intended zero point ten after I had cloned a chart. It takes longer to describe than to do, and it is the only reason I trust a restart.